1/*
2 *  linux/arch/arm/lib/memset.S
3 *
4 *  Copyright (C) 1995-2000 Russell King
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
9 *
10 *  ASM optimised string functions
11 */
12#include "assembler.h"
13
14	.text
15	.align	5
16
17ENTRY(memset)
18	ands	r3, r0, #3		@ 1 unaligned?
19	mov	ip, r0			@ preserve r0 as return value
20	bne	6f			@ 1
21/*
22 * we know that the pointer in ip is aligned to a word boundary.
23 */
241:	orr	r1, r1, r1, lsl #8
25	orr	r1, r1, r1, lsl #16
26	mov	r3, r1
27	cmp	r2, #16
28	blt	4f
29
30#if ! CALGN(1)+0
31
32/*
33 * We need 2 extra registers for this loop - use r8 and the LR
34 */
35	stmfd	sp!, {r8, lr}
36	mov	r8, r1
37	mov	lr, r1
38
392:	subs	r2, r2, #64
40	stmgeia	ip!, {r1, r3, r8, lr}	@ 64 bytes at a time.
41	stmgeia	ip!, {r1, r3, r8, lr}
42	stmgeia	ip!, {r1, r3, r8, lr}
43	stmgeia	ip!, {r1, r3, r8, lr}
44	bgt	2b
45	ldmeqfd	sp!, {r8, pc}		@ Now <64 bytes to go.
46/*
47 * No need to correct the count; we're only testing bits from now on
48 */
49	tst	r2, #32
50	stmneia	ip!, {r1, r3, r8, lr}
51	stmneia	ip!, {r1, r3, r8, lr}
52	tst	r2, #16
53	stmneia	ip!, {r1, r3, r8, lr}
54	ldmfd	sp!, {r8, lr}
55
56#else
57
58/*
59 * This version aligns the destination pointer in order to write
60 * whole cache lines at once.
61 */
62
63	stmfd	sp!, {r4-r8, lr}
64	mov	r4, r1
65	mov	r5, r1
66	mov	r6, r1
67	mov	r7, r1
68	mov	r8, r1
69	mov	lr, r1
70
71	cmp	r2, #96
72	tstgt	ip, #31
73	ble	3f
74
75	and	r8, ip, #31
76	rsb	r8, r8, #32
77	sub	r2, r2, r8
78	movs	r8, r8, lsl #(32 - 4)
79	stmcsia	ip!, {r4, r5, r6, r7}
80	stmmiia	ip!, {r4, r5}
81	tst	r8, #(1 << 30)
82	mov	r8, r1
83	strne	r1, [ip], #4
84
853:	subs	r2, r2, #64
86	stmgeia	ip!, {r1, r3-r8, lr}
87	stmgeia	ip!, {r1, r3-r8, lr}
88	bgt	3b
89	ldmeqfd	sp!, {r4-r8, pc}
90
91	tst	r2, #32
92	stmneia	ip!, {r1, r3-r8, lr}
93	tst	r2, #16
94	stmneia	ip!, {r4-r7}
95	ldmfd	sp!, {r4-r8, lr}
96
97#endif
98
994:	tst	r2, #8
100	stmneia	ip!, {r1, r3}
101	tst	r2, #4
102	strne	r1, [ip], #4
103/*
104 * When we get here, we've got less than 4 bytes to zero.  We
105 * may have an unaligned pointer as well.
106 */
1075:	tst	r2, #2
108	strneb	r1, [ip], #1
109	strneb	r1, [ip], #1
110	tst	r2, #1
111	strneb	r1, [ip], #1
112	mov	pc, lr
113
1146:	subs	r2, r2, #4		@ 1 do we have enough
115	blt	5b			@ 1 bytes to align with?
116	cmp	r3, #2			@ 1
117	strltb	r1, [ip], #1		@ 1
118	strleb	r1, [ip], #1		@ 1
119	strb	r1, [ip], #1		@ 1
120	add	r2, r2, r3		@ 1 (r2 = r2 - (4 - r3))
121	b	1b
122ENDPROC(memset)
123